LetThemBuild launches multi-model AI code review tool
The new desktop application allows a lead AI to perform tasks while multiple other models verify the work in real-time.

The short version
- Supports multiple models including Claude, ChatGPT, Gemini, and Grok out of the box.
- Provides a 'receipt' for every review showing the exact commands run and their exit codes.
- Works on macOS, Windows, and Linux using a git worktree for local development.
- Allows users to steer the AI process from any device, including iPhones and web browsers.
Why it matters
It moves AI coding from a single-model 'black box' to a verifiable group process. This helps developers ensure that AI-generated code actually works by requiring multiple models to prove their checks.
A collaborative AI workspace
LetThemBuild is a new desktop application designed to facilitate group code reviews using various artificial intelligence models. The platform allows a lead model to perform a specific task while three other models review that work simultaneously. This setup ensures that the output is checked by multiple perspectives before it is finalized.
Users can choose from several popular models like Claude, ChatGPT, Gemini, and Grok directly from the app. For those who prefer specific configurations, the tool also supports connecting to hundreds of other models through custom endpoints. This flexibility allows teams to use the specific AI tools they already pay for or maintain.
The system is designed to work across different operating systems, including macOS, Windows, and Linux. It functions by creating a git worktree, which is a way to have multiple working directories for a single repository. This allows the AI to perform work in a dedicated space without disrupting the main project.
Verifiable checks and receipts
One of the core features of the tool is the requirement for verifiable evidence during the review process. A reviewer model cannot simply say that the code looks good; it must actually run checks to provide proof. The app generates a receipt for every review that lists the specific commands the reviewer ran and their resulting exit codes.
This transparency means that a reviewer who fails to run any tests is clearly marked as having run nothing. This prevents the system from providing false positives where a model might agree with a change without actually testing it. The 'warden' role in the system monitors every action taken by the models to ensure consistency.
The final output of this process is a branch that can be pulled back into the main project. Once the group of models reaches an agreement, the work can be turned into a pull request. This workflow mimics a standard human development cycle but replaces some of the manual checking with automated, multi-model verification.
Remote steering and cross-device use
While the main work happens on a desktop application, LetThemBuild allows users to steer the process from almost any device. A user can hand a task to the cloud from a Mac, a web browser, or even an iPhone. This means the primary computer can stay closed while the user monitors the progress from a mobile device.
The companion web and iPhone app, called LetThemChat, shares the same account and sessions as the desktop tool. These mobile interfaces are intended for reading what the models have done, asking the group questions, and directing the next steps. This creates a cohesive experience where the developer is not tied to a single screen.
The tool also offers different intensity levels for how the models work. Users can choose 'Quick' for simple questions, 'Consensus' for changes that require agreement from everyone, or 'Deep' for tasks that must be right. There is also a 'Default' mode where the room decides the level of effort based on the request.
Privacy and local control
Security and data privacy are central to the design of the LetThemBuild platform. The application explicitly states that nothing leaves the user's machine unless the user chooses to allow it. The app provides a clear warning before any data is sent to the cloud, ensuring the user remains in control of their information.
Users have the option to use the keys provided by the developers or bring their own API keys. By using their own keys, developers can keep their data within their existing subscription frameworks. This choice allows for a balance between ease of use and strict data sovereignty for sensitive projects.
The documentation for the product is publicly available without requiring a sign-up. It covers the mechanics of how the room reaches an agreement and what the app is allowed to touch on the user's machine. This transparency is intended to help users understand the boundaries of the AI's permissions.
The path to agreement
The workflow is structured to move from a single prompt to a finalized piece of code. The lead model builds the feature, and the reviewers run checks to verify the logic. Only when the reviewers provide evidence of success does the work count as done, allowing the user to move it back to their local machine.
This multi-layered approach addresses the common issue of AI hallucination or incorrect logic in complex coding tasks. By forcing multiple models to act as a check on one another, the system reduces the likelihood of errors reaching the final codebase. It turns the AI interaction into a structured committee rather than a single conversation.
Moving forward, the tool aims to streamline how groups of models interact with local filesystems. By providing a clear receipt of every action, it creates a trail of accountability that is often missing in standard AI coding assistants. This makes it a viable tool for developers who need to verify the integrity of every line of code generated by an AI.
What to watch
- How many different models can be connected via the OpenAI-compatible endpoint?
- Whether the 'Deep' mode significantly increases the time it takes to complete a task.
Try it
Check the LetThemBuild documentation to see the specific rules for how the AI room reaches agreement.
Sources and links
- OfficialOpen LetThemChat letthemchat.com
- Reportingletthembuild.com: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by letthembuild.com · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
